Strata 0.2.14 has been released, now utilizing self-written GPU kernels for prompt processing, which eliminates its last dependency on llama.cpp, resulting in improved performance. The update shows notable speed increases, particularly for specific prompts on AMD's Strix Halo, where processing time dropped significantly. Despite these advancements, some skeptics raise concerns about the quality of claims, inconsistencies in documentation, and a substantial percentage of downloads being attributed to the demo trailer. Strata continues to operate with a limited number of models to enhance speed and maintains a stable OpenAI-compatible serving layer while treating the runtime as interchangeable

Strata 0.2.14 Ships Self-Written Prompt Kernels, Drops llama.cpp Dependency

Strata 0.2.14 now reads prompts using GPU kernels it wrote itself, shedding its last hard dependency on llama.cpp. The change delivers mostly single-digit speed gains across hardware, with the standout being a 256-token prompt on AMD's Strix Halo dropping from 2,010 ms to 408 ms. Skeptics point out that quality claims remain qualitative, documentation contradicts itself, and one audit found roughly 19% of downloads were just the demo trailer. Strata runs only three models in exchange for speed, and the recommended setup keeps a stable OpenAI-compatible serving layer while treating the runtime itself as replaceable.

Strata 0.2.14 Ships Self-Written Prompt Kernels, Drops llama.cpp Dependency @ Linux Compatible